About Me
Zili is an aspiring AI safety researcher focusing on evaluations. She is currently exploring alignment issues with LLM agents carrying out long-horizon tasks. She graduated from Yale in 2025 with a Ph.D. in astrophysics. For her thesis project, she led the science analysis of the Dragonfly Ultrawide Survey and built a custom pipeline on AWS.
Apart from research, she writes about science. She published articles on Astrobites and worked at the Yale Graduate Writing Lab as a writing consultant.
AI Research
Ongoing Project: Evaluating agents on long-horizon tasks
A project funded by the Algoverse AI Safety Fellowship.
LLM agents are being deployed to carry out tasks on behalf of their users. Some of those tasks involve many steps and a large context. During this process, are agents able to retain their alignment training, remember the original goal, and deal with potential changes in the environment that they were not trained on?
Astrophysics
Publications
Ph.D. Thesis Project: The Dragonfly Ultrawide Survey
Shen, Z. et al. 2024, The Astrophysical Journal, 976, 75
- Our research group built the Dragonfly Telephoto Array to detect very large and diffuse galaxies that might be missed with conventional telescopes.
- We used this telescope to map 10,000 square degrees of the northern sky.
- I led the science analysis of this entire dataset to find new galaxies.
- I built a custom data pipeline on AWS and secured funding ($10K) to deploy it.
- We found 11 large, low surface brightness galaxies with spectroscopic confirmation.
Press and Outreach:
- I won third place at the Yale Three-minute Thesis Competition, competing against Ph.D. students from all over Yale to present my thesis work in just three minutes.
- The background image of this website comes from data I obtained using the Hubble Space Telescope. I used this data to measure the distance to a galaxy lacking dark matter. The results been reported by STScI, Yale, and IAS.
Writing
